A facial recognition system is a system that can automatically identify a person by matching a live image of the person to a reference image.
FIG. 1 illustrates a conventional facial recognition system 100 including a camera 102, a reference image database 104, and a comparator 106. Some reference images are stored in the reference image database 104. The camera 102 captures a live image of a user and forwards the live image to the comparator 106. The comparator 106 compares the live image with the reference images in the reference image database 104, and generates a recognition result indicating whether the live image can match to a reference image.
This conventional facial recognition system may perform well when the current environment (the lighting and direction of lighting, etc.) is similar to the environment in which the reference image is taken. However, when the current environment is different from the environment in which the reference image is taken, the performance of the facial recognition system can drop significantly and the live image of the user may not match to any reference image in the database, which may lead to a high possibility of recognition failure.